Upload
others
View
8
Download
0
Embed Size (px)
Citation preview
Multi-ray photogrammetry: A rich dataset for the
extraction of roof geometry for 3D
reconstruction
Andrew McClune, Pauline Miller, Jon Mills
Newcastle University
David Holland
Ordnance Survey
Background
Approaches used in latest ISPRS benchmark on 3D reconstruction:
• 10 methods using original ALS
• 1 method using DSM from ALS
• 2 methods using imagery
• 1 method combined ALS and imagery (Rottensteiner et al, 2014)
Xiong et al, 2014
Multi-ray photogrammetry
Haala, 2014
1 image
15 images
Analogue camera overlap
Digital camera overlap
Number of images per pixel
Aim & Objectives
The aim of this project is to use data produced from multi-ray photogrammetry to automatically extract and reconstruct LoD 2 3D building models, with accurate geometry of roof vertices:
• Investigate current methods used for building modelling and rooftop extraction, examining their strengths and weaknesses;
• Develop an automated procedure for the extraction of 2D and 3D rooftop vertices from the datasets created from multi-ray photogrammetry;
• Integrate and refine the extracted data from Objective 2 to create a 3D roof model and then integrate with a LoD 1 building model to create the final LoD 2 building model.
• Validate and refine the developed methodology, delivering appropriate metrics on accuracy and reliability.
Test Site
• Newcastle upon Tyne, UK
• Imagery captured using
UltraCam XP camera in
November 2010,
• Data processed in UltraMap
3.0
• Raster DSM
• Image based point cloud
• True Orthophoto
Workflow & Test Site
2D: Edge Detection
• Building extracted using OS Mastermap data.
• Investigation of common edge detectors
• Canny edge detector to extract perimeter of roof, ridge and valley lines
• Also extracts noise from roof texture and shadow
Roof extracted from orthophoto Canny edge detection
3D: Scan line segmentation
• Originally proposed by Jiang & Bunke (1994) for segmentation of close range imagery.
• Classify cross sections of an image into lines then perform region growing to merge neighbouring lines and form planes.
• Investigated by Alharthy & Bethel (2004) for 3D reconstruction of buildings from lidar
• Limitations from point cloud spacing and polygon extraction moving the nodes from true position.
Alharthy & Bethel, 2014
Scan line segmentation – Gable roof
• Detection of breakpoints based on a change in the sign of the slope
Ortho
extraction
nDSM
profile
Scan line
segmentation
Profile of
detected
breakpoints
Scan line segmentation – M-Shaped roof
• Detection of breakpoints based on a change in the sign of the slope
Ortho
extraction
nDSM
profile
Scan line
segmentation
Profile of
detected
breakpoints
Integration of datasets
• Canny edge detection extracts accurate location of roof edges but also extracts false positives from roof planes
• Scan line segmentation can be used as part of a process to extract planes but does not extract accurate location of roof edges
Scan line segmentation with edge detection
• Linear Regression of detected Canny edges incorporated into
• Iteratively fit line through data points and remove edges based on residuals of the fit
Scan line segmentation with edge detection
• Linear Regression of detected Canny edges
• Iteratively fit line through data points and remove edges based on residuals of the fit
Canny edge detection Roof breakpoints from scan line
segmentation
Conclusions & Future Work
Conclusions
• Multi-ray photogrammetry offers an abundance of dense data that can be used for roof geometry extraction;
• Canny edge detection and scan line segmentation have their individual strengths which can overcome weaknesses for extracting roof geometry;
Future Work
• Exploring threshold sensitivity of canny edge detection and linear regression
• Vectorising of extracted breakpoints
• Developing scan line segmentation approach for removal of vegetation
Thank you for listening
ReferencesAlharthy, A. and Bethel, J. (2004) 'Detailed building reconstruction from airborne laser data using a moving surface method', International Archives of Photogrammetry and Remote Sensing, 35, pp. 213 - 218.
Haala, N. (2014) ‘High density image matching’ EuroSDR EduServ 2014 Pre-course seminar, Lecture conducted from Trento, Italy.
Jiang, X. and Bunke, H. (1994) 'Fast segmentation of range images into planar regions by scan line grouping', Machine Vision and Application, (7), pp. 115-122.
Rottensteiner, F., Sohn, G., Gerke, M., Wegner, J.D., Breitkopf, U. and Jung, J. (2014) 'Results of the ISPRS benchmark on urban object detection and 3D building reconstruction', ISPRS Journal of Photogrammetry and Remote Sensing, 93(0), pp. 256-271.
Xiong, B., Oude Elbrink, S., Vosselman, G., (2014) A graph edit dictionary for correcting errors in roof topology graphs reconstructed from point clouds. ISPRS Journal of Photogrammetry and Remote Sensing 93, 227-242.
AcknowledgementsThis project is part funded by the Ordnance Survey and ESPRC. The authors would like to thank Ordnance Survey for the
provision of aerial imagery, photogrammetric point cloud and DSM.